Upkeep

In the recently remastered WarCraft III (WC3) we can find the upkeep mechanic. It's often perceived negatively. In this post I'll try to explore it as game-design element, see how it affects various game dynamics, and whether it could be in theory replaced by other mechanics.

How it works

• If your supply is high, you gain only a certain fraction from the collected primary resource (gold)
• 0–50 supply: 10 gold per worker trip
• 51–80 supply: 7 gold per worker trip
• 81–100 supply: 4 gold per worker trip
• Gold mines always lose 10 gold per worker trip even if you are getting less gold
• Gold from other sources is reduced as well (despite it not being indicated)

FRAMEWORK

I'll be looking at this through MDA framework. To simplify what this framework is about: mechanics are for example rules of the game and something game-designers can change directly. Dynamics are how games play out. And aesthetics are emotional responses evoked in players. The causal chain will usually look like this:

Mechanics (M) → Dynamics (D) → Aesthetics (A)

Rules lead to gameplay, and the gameplay evokes some feelings in players. However, in the case of upkeep and in the next example, there is also a direct effect of mechanics on aesthetics, which makes this an interesting problem to explore.

Mechanics → Aesthetics

Experience in World of Warcraft

In World of Warcraft (WoW), the experience gain for players was originally halved after some time. This was received very poorly in user testing. Eventually, the system was changed to include "rest experience" instead, which provided limited bonus experience, and was seen much better by players.

"From a purely numerical standpoint, it didn't really affect how fast you were leveling, but it had a huge psychological effect on how people thought about the system,"

"As a game designer it was a great learning experience for me and I made it actually one of the game design values of the studio: Make it a bonus. Whenever you're trying to prevent players from doing something in your game, ask the question 'Is there a way to create a bonus to do the opposite behavior?'"

– Rob Pardo
Game designer on both WoW and WC3

This example shows that there can exist distinct mechanics that lead to the same dynamics while also having vastly different outcomes on aesthetics.

M₁ → D₀ → A₀   |   M₁ → A₁

M₂ → D₀ → A₀   |   M₂ → A₂

In this case A₀ is base reaction on having diminishing returns on gameplay length; A₁ is a negative reception of perceived experience loss; and A₂ a positive reception of perceived experience gain. This is a result of irrational workings of human mind – namely loss aversion and framing that can make humans perceive numerically identical situations in very different ways.

Unfortunately for WC3, the quote is regarding WoW development which came few years after WC3 development and the upkeep mechanic was already in place. WC3 was released in 2002 and WoW in 2004.

Approach

The situation with upkeep is similar to that with the experience in World of Warcraft. There is some dynamic that we want, but the mechanic itself has a direct negative impact on aesthetics (A₁). In WoW the solution was as simple as reversing the mechanic, which led to an identical dynamic but with a positive direct effect on aesthetics (A₂).

In theory, you could do something similar in the case of upkeep – giving players more resources if they have fewer units. However, that wouldn't make sense for most RTS games. Instead, I'll try to go over all dynamics that the upkeep affects, and see if there are other mechanics that can replace it in each situation.

To keep this simple I will stay close to WC3 design and fantasy when considering options, even if the design space for all RTS games is bigger than that.

Issues with upkeep:

• It feels punishing due to loss aversion.
  
• Arbitrary cutoffs (50, 80, 5, 7) are unintuitive and immersion breaking. At one moment you have full income, and suddenly you are losing 30% of it.
• The same can be said about not having to "pay your army" as long as you are not mining anything. This also sends mixed signals to players and encourages them to stop workers from mining gold.
• Numbers from selling items are not correctly shown when they are reduced by upkeep.

Upkeep's effects following game dynamics:

1. Snowballing
2. Expanding
3. Soft cap
4. Past and future cost
5. Depth
6. Economic tension

With each dynamic I'll look at how the upkeep mechanic affects it and go over some possible mechanics that could have a similar effect ingame.

1. Effect on snowballing

Upkeep reduces snowballing – it serves as a negative feedback loop. A player with lower supply will have easier time to catch up. For example a player on one base gets additional 180 gold each minute with no upkeep compared to being on low upkeep. That can result in around one additional unit every minute as long as the player is on one upkeep tier lower.

~ ~ ~

This might be the best argument for upkeep as there is not a simple and fitting mechanic with the exactly same result. I have written about feedback loops in StarCraft II, and in WC3 you could introduce new or empower already existing negative feedback loops to step in upkeep's role:

• Bonus experience / bounty when having fewer units
• Various forms of defender's advantage
• Worse scaling with unit numbers, e.g.:
• Melee units don't scale well with numbers
• Offensive area-of-effect abilities scale with enemy numbers
• Farms – investment into increasing supply makes it easier to catch up and harder to gain supply advantage. For example in StarCraft II rebuilding a lost Zealot costs 100 minerals, but building a new one with the supply cost added is 125 minerals.
• Mechanical difficulty of controlling armies in itself makes bigger armies less effective. This is especially pronounced in Brood War with its strict unit and structure selection limit, no smart cast and non-optimal pathing.
• Other options that wouldn't easily fit WC3:
• Partial refund for lost units. This might fit Undead but not all factions.
• Other buffs gained from losing units.
• Changing unit cost depending on the number of currently living or lost units.

High-skill ceiling and diverse unit roles can to some degree work in a similar way by letting fewer units be still effective – it's not just a game of numbers and doesn't necessary snowball because of having few units more.

2. Effect on Expanding

Upkeep disincentives expanding as with more workers you are likely to reach higher upkeep tiers faster. However, that doesn't mean expanding is never a good choice, with an additional base and 5 more workers, the player gets +40% gold income even when on one upkeep tier higher. Further expanding on three and more bases is heavily discouraged.

~ ~ ~

Tweaking expansion investment cost, safety, and return rate plus additional supply cap tweaks could lead to the same dynamics here. Although expanding isn't as important and common as in other RTS games like StarCraft I & II.

The fact that 10 gold is lost from the Gold mines per worker trip no matter what means the main gold mine will mine out at the same time for both players. This will be around 19 minutes in normal games. Upkeep has no effect on this. From what I found average length of solo game is around 15.8 minutes, but still a decent number of games will go over 19 minute mark.

Distribution of games by their lengths [minutes] (source)

3. SOFT CAP

Upkeep serves as a soft supply cap by providing diminishing returns on army resource investment. This adds strategic decisions to the game, something that would be not there with just lower supply cap. There is a choice how far will you push this soft cap. It can sharpen timings and strategies. Fixed supply limit cannot be pushed.

~ ~ ~

• Optional global or unit upgrades can provide the same diminishing returns on resource investments.
• More temporary power-ups or upgrades that cost resources can help to sharpen timings and strategies as well. Purchasable scrolls and other consumables work in this way.

4. Past and future cost

One difference between upkeep and the previous examples is where the cost lies. If you bought a unit, item or a consumable the cost is in the past, and now it's time to make use of it. In case of upkeep the cost starts to take effect when you step over an upkeep threshold and lasts until a battle.

That cost in future mining is sharpening your current offensive timing. It will add a bit more strength to your push, but the long term play will suffer for it. This is similar to not expanding behind your push in StarCraft II, but upkeep enables to have something like that despite both players being on one base.

~ ~ ~

Possible replacements with a similar effect:

• Pulling workers to bigger push
• More options for mid and long-term progression – upgrades and tech
• More ways how to invest into the economy – expansions, workers, mining upgrades
• Some sacrifice mechanic for workers or resources to gain a temporary power bonus. That's equivalent to worker pulling only the implementation is different.

5. Depth

Depth isn't a game dynamic, but you could say it's a property of game dynamics. Upkeep definitely adds some depth to the game, and managing upkeep is a skill and a strategic play on its own.

~ ~ ~

But it's not a question whether to have depth or not, it's where and how you add depth to a game. In the case of upkeep, I would argue that its cost in aesthetics is too high. If you want to just add depth to the economic part of the game you will likely find better options than upkeep.

6. Economic tension

The reason given by Rob Pardo on Designer Notes (~1h:44m).

"When upkeep wasn't in there, then all you really had to control the players was a population cap. But what would happen... what was always happening in playtests is people would get up to their pop cap, and you know they would have a big army, and then they would just stockpile the gold, they would have a ton of gold. And then what would happen is, we would end up in combat with each other, and if it did end up not resolving the war, you know, both of us could basically instantly rebuild our army. There was just no economic tension."

– Rob Pardo
Game designer on both WoW and WC3

~ ~ ~

I was surprised this was the argument he gave. In my opinion the anti-snowballing argument (1) is a more compelling one. What he is describing is a lategame crisis which many RTS games have to deal with, and have done so in many ways. WC3 also has some unique options to force player being active on the map – namely commander experience, loot, and control neutral shops and other points of interest.

In my view focusing on having a healthy and long mid-game is always better than coming up with bandaid solutions for lategame. Economy should be important, engagement should happen often, units should be lost. Players shouldn't often max-out while stockpiling enough resource to max-out again. If the mid-game is still too short, then look at the economy and tech progression. Force players to be active on the map through rewards (XP, loot, points of interest) and economy (expanding, collecting). I won't go into details here, as this is the core of the game, what motivates players, what is the structure of goals ingame. This will be different between games.

Now let's say the game's early and mid-game are working great, and 70-80% games end there. And they should end there. If you characterize lategame as having enough money to reach your desired unit composition while also having enough resources to rebuild it "instantly", then that's inherently less interesting state than such where you have to think where you invest your resources. Thus staying in "mid-game" is better, and you should have only a smaller fraction of games in this type of lategame to spice up the diversity of games. This is something to consider when thinking about changes targeting lategame.

Again, early and mid-game are good, what about lategame? There are various options to deal with the issue of quick remaxing and lacking tension.

• Remaxing takes time, especially for better units. In the meantime you might lose your expansions, strategic positions or even the game.
• Units reliant on energy/mana take time to reach their full strength and can be crucial for lategame.
• Remaxing takes a lot of resources – especially for good units. Consumables might be limited or costly, the same for units purchased from neutral shops.
• Restricted economy and more options for other resource sinks – static, consumables, etc.
• In all previous examples asymmetric faction design can add even more tension. Zerg in StarCraft might remax more quickly, but its expansions are also less defended, Zerg probably lost more and traded less efficiently whole game. This introduces more timings with tension where players have advantages over each other.
• Games like C&C3 sidesteps lategame issues due to how the economy scales – reaching maximum economy quickly and from there the income decreases.

~ ~ ~

Overall in this particular argument the upkeep is a bandaid solution for lategame. In my opinion a broader look at game design of all stages of the game (early, mid and lategame) would be better.

Company of heroes

This post is mostly about WC3, but I have to mention Company of Heroes as it's another game with upkeep. Its implementation is different and works rather well. That's partly because the economy system is very different compared to games like WarCraft or StarCraft.

To simplify:

• In Company of Heroes resources are generated constantly and automatically
• Every unit costs some resources to deploy, and then reduces the main resource income (manpower) by a small amount.

In this case the income isn't decreased by arbitrary percents at arbitrary breakpoints, instead each unit reduces the income by a fixed amount. It's more intuitive and usually doesn't send mixed signals to players.

Overall this implementation works a lot better when it to comes to the effect on aesthetics. The effect on game dynamics would be hard to compare as they are were different games.

my take on upkeep

I wanted to look at it because it's a very interesting design problem. Upkeep has some things going for it, and its effect on certain game dynamics is very positive. That said, I don't think using upkeep in WC3's implementation was a good choice then, and it would be even worse decision today. And yet other implementations of upkeep as seen in Company of Heroes can work quite well.

There isn't a good full replacement for its anti-snowball effect, where you can get few more units if you are behind on army supply. However other games managed to lean on other negative feedback loops. Other games might not be as hero-centric, but hero focus can both dampen or empower snowballing depending on its implementation.

Other than snowballing, there are mechanics that can be used instead of upkeep to accomplish similar results. Either one of them might or might not fit a particular game well.

~ ~ ~

Thank you for reading this post. It has been a bit more difficult to write this one. I didn't want to sound overly negative, and I hope it's been interesting even if you do not share the same opinion as me.

Links to check out

• https://www.idlethumbs.net/designernotes/episodes/rob-pardo-part-1
  (WarCraft III's upkeep mentioned at 1h:44m)
• Grubby's video on upkeep (some nice arguments for it)
• https://www.usgamer.net/articles/how-world-of-warcraft-was-made-the-inside-story
• https://warcraft3.info/articles/177/battlenet-statistics-part-1-how-many-active-players-are-there-in-2018
• https://warcraft3.info/articles/238/new-to-warcraft-upkeep-taxes
• https://waywardstrategy.com/2020/07/06/anti-snowball-design/
• https://www.maguro.one/2019/01/feedback-loops-in-starcraft